Open Source · PyPI · Python 3.11+

Unified Nexus

One decorator. FastAPI REST and MCP for AI agents — simultaneously. No drift. No duplication. One source of truth.

Get Started → ★ View on GitHub

pip install unified-nexus v0.1.0 · MIT

The Problem

Stop writing the same logic twice.

Traditional stacks force you to define a FastAPI route, then manually re-wrap it as an MCP tool. Over time they drift apart — different validation, different behavior. Unified Nexus ends that permanently.

❌ Traditional Stack

# Two files. Two definitions.
# They will drift. Always.

routes/user.py        ← HTTP logic
mcp_tools/user.py     ← MCP copy-paste
models/user.py        ← maybe shared?

# Result: Code Drift. Silent bugs.
# API validates X. Tool validates Y.

✅ Unified Nexus

# One file. One decorator.
# Always in sync. Guaranteed.

main.py               ← REST + MCP
@nexus.universal_tool ← both at once
                      ← that's it.

# Result: One source of truth.
# HTTP and AI share exact same logic.

Why Unified Nexus

Everything you need.
Nothing you don't.

⚡

Single Source of Truth

Define logic once. HTTP and MCP interfaces share identical Pydantic validation, schemas, and business logic — forever in sync.

🤖

AI-Native by Default

Python docstrings become rich MCP tool descriptions. Field annotations guide Claude, Cursor, and other LLM hosts on how to call your tools correctly.

🔍

Auto-Method Detection

Intelligently selects GET or POST based on your function signature. Pydantic input → POST. Query params → GET. Zero config.

🛡️

Shared Middleware

Apply auth, rate limiting, or logging once. The same middleware protects both your HTTP API and MCP interface — no duplication ever.

💬

LLM-Optimized Errors

Python exceptions are converted to natural-language hints that help LLMs self-correct their tool calls instead of returning raw stack traces.

🔒

Lifespan-Safe MCP

Correctly composes FastAPI and FastMCP lifespans. The MCP task group is always initialized before requests arrive — no runtime surprises.

Quick Start

Up and running in 60 seconds.

Install, define a tool, run. Your function is immediately available via both REST and MCP.

# Install via pip
pip install unified-nexus

# Or clone from GitHub
git clone https://github.com/rajboopathiking/UnifiedNexus
cd UnifiedNexus
pip install -e .

from unified_nexus import UnifiedNexus
from pydantic import BaseModel, Field

nexus = UnifiedNexus("MyNexusAPI")

class UserRequest(BaseModel):
    user_id: int = Field(..., description="The unique ID of the user")

@nexus.universal_tool(path="/user-info")
def get_user(req: UserRequest):
    """
    Fetches user status and clearance level.
    This docstring is used by Swagger AND AI agents!
    """
    return {"id": req.user_id, "status": "Active", "tier": "Gold"}

app = nexus.finalize()

if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=8000)

# Start the server
python main.py

# ✅ REST API Docs
http://localhost:8000/docs

# ✅ MCP Endpoint
http://localhost:8000/mcp

# ✅ Verify MCP with Inspector
npx -y @modelcontextprotocol/inspector http://localhost:8000/mcp

# Async tools work natively
@nexus.universal_tool(path="/fetch")
async def fetch_data(resource_id: str):
    """Fetches data asynchronously from external source."""
    async with httpx.AsyncClient() as client:
        res = await client.get(f"https://api.example.com/{resource_id}")
    return res.json()

# Custom path + explicit method
@nexus.universal_tool(path="/search", methods=["GET"])
def search(query: str, limit: int = 10):
    """Search the catalog. Returns ranked results."""
    return {"results": [], "total": 0}

AI Integrations

Connect any AI agent in seconds.

Your MCP endpoint works with every major AI host out of the box — no extra config, no plugins.

🟣 Claude Desktop

{
  "mcpServers": {
    "my-api": {
      "url": "http://localhost:8000/mcp",
      "transport": "streamable-http"
    }
  }
}

🔵 Cursor / VS Code

# Add to MCP settings:

Server URL:
http://localhost:8000/mcp

Transport:
streamable-http

# Your tools appear in
# the AI sidebar instantly.

🔍 MCP Inspector

# Test & debug all tools:

npx -y @modelcontextprotocol/inspector \
  http://localhost:8000/mcp

# Opens at localhost:6274
# Full tool call UI included.